The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The README.md and requirements.txt suggest installing a tool named search-cli from a third-party Homebrew tap (199-biotechnologies/tap). This external tool is the primary mechanism for data retrieval and operates outside the standard package registries.
[COMMAND_EXECUTION]: The skill uses the Bash tool and subprocess.run to execute various commands, including search-cli, weasyprint for PDF generation, and several internal Python scripts (citation_manager.py, evidence_store.py, etc.) to maintain research state and validate output.
[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Category 8). It ingests untrusted data from the web (Phase 3: RETRIEVE) and interpolates this data into prompts for synthesis and critique (Phases 5 and 6). While the skill employs structured evidence capture and validation scripts, the synthesis of external content remains an inherent risk factor for prompt-based manipulation.

deep-research